Blur Busters Forums

Hi my name is Chris I am a long time lurker.
I like to share this idea/concept with you.
If we move Framegen into the monitor we could have a 4k 600hz to 1200hz monitor very soon without addional bandwith requirements.

Please feel free to share your thoughts

Chris, in-panel framegen could fake 600–1200 Hz without more bandwidth, but adds latency and artifacts. True 4K @ 1000 Hz needs major panel and link advances, so we’ll see smarter tricks and ~500 Hz panels first.

It's an interesting concept; I wrote about display-side co-GPU concept about ten years ago in one of my articles, so this is not a new idea.

In-monitor framegen will require a lot of new protocols to communicate lots of ground truth from the computer to the monitor, to reduce the number of artifacts. Black box interpolation is laggy, so you need to throw a lot of ground truth at framegen to remove the latency from it.

Some precedent: Some standalone VR headsets can do during PCVR streaming mode. For example WiFi streaming or computer-side underperformancce. For example, Quest 2/3 VR headset will do headset-side warping at 90fps even if the AirLink is streaming at a low framerate or the computer-side GPU is running at a lower framerate.

Now that said, 16K 60 Hz = 8K 240 Hz = 4K 960Hz in pixels/sec bandwidth. With maximum DSC, it kind of fits into the new DP 2.1 protocol. It may, be cheaper to manufacture OLEDs that do 4K 1000Hz via traditional means. And doing the co-GPU concept as a 2-thread instead on the PC side (to keep the commoditized display ecosystem -- otherwise, displays start hitting four or five figure prices).

However, 2-thread approaches (whether PC-side GPU or display-side co-GPU, or the concept below), is one way to make framegen perceptually lagless, at least for mouselook-style situations.



Now, that being said, more displays are getting GPUs built in. The Blur Busters Open Source Display Initiative, www.blurbusters.com/open-source-display -- touches upon the potential of using that GPU for various purposes including blurbusting duty. That can include things like simple 3dof reprojection/scrolls/turns/pans/etc. That would allow extra framerate, if done via a kind of an API transport superimposed on the display channel, to provide extra temporal resolution data -- not too dissimilar to this idea (but on a much lighter weight co-GPU basis, like Quest 2/3 headset-side 3dof for PCVR underperformance)

This idea (in a roundabout way) is being considered as one of a large toolbox of potential solutions, offered as a user choice.

Chief Blur Buster wrote: ↑

14 Aug 2025, 13:04

It's an interesting concept; I wrote about display-side co-GPU concept about ten years ago in one of my articles, so this is not a new idea.

Yes, your articles are incredibly detailed with so many ideas/concepts it´s sometimes difficult to keep track of the information on blur busters, it could be that I read about it the past. If so sorry, I don´t want to steal your ideas.

@natalie I understand many people don´t like Framegen much, due to latency but one could also go like FSR in the monitor.

Here is a small mini specc:


In-Monitor Upscale + Frame-Gen (IMUF) — Mini-Spec v0.1

Goal: Drive a 4K panel at 500–1000 Hz with perceptually lagless motion using commodity links by transporting a lighter base stream (e.g., 1440p@240–300 Hz) and doing FSR-style upscaling + ×2–×4 reprojection in the monitor, with optional BFI.

1) Transport & Bandwidth

Performance Transport Mode (PTM): GPU sends a lower-res, high-Hz stream over DP 2.1/HDMI 2.1; monitor advertises capability via EDID vendor block (see §4).

Relative pixel-rate table (vs 4K@250 Hz = 100%)

Base stream Relative pixels/sec Notes
4K@250 Hz 100% Reference (too heavy without DSC)
1440p@250 Hz 44% Sweet spot for quality/latency
1440p@300 Hz 53% Extra headroom for reprojection
1080p@360 Hz 35% Budget GPUs / esports focus

DSC may be used on the link; PTM does not require DSC to function.

2) In-Monitor Processing Pipeline
GPU MONITOR (SoC/NPU/DSP) PANEL
Base frames @ PTM ──► [Line buf] ─► Upscale (FSR-like) ─► Reproject ×2–×4 ─► OD/BFI ─► Scanout @ 500–1000 Hz
(e.g., 1440p@300) ~0.1ms 0.3–0.7ms 0.3–0.6ms 0.05–0.1ms (rolling)


Latency budget (target end-to-end added): ≤ ~1.2 ms

Line-buffered upscale to avoid full-frame queues.

Reprojection fed by motion vectors + (optional) depth/occlusion mask when available; otherwise fall back to optical flow.

BFI (optional): single-strobe per generated frame; PWM-free; user-tunable duty.

Artifact plan

Disocclusions/ghosting: clamp vectors at depth edges; blend from native samples; hole-fill with spatial taps.

Overdrive sync: OD solver consumes generated subframes to prevent smear.

VRR: Base stream remains VRR; reprojection smooths between intervals.

3) Sideband “Ground-Truth” Metadata (optional, recommended)

Per-frame packet (sent alongside base frame):

MV grid: 8×8 or 16×16 macroblocks, (vx, vy) as 16-bit fixed-point each.

Depth/occlusion mask: 8-bit, same grid (0–255 = near→far).

Exposure/jitter info: for TAA-style recon.

Frame ID & timestamps: 32-bit each (sync/latency calc).

Transport options (vendor-agnostic):

DP secondary data channel / sideband payload;

HDMI infoframe extension;

AUX/I²C mailbox for capability + small state.

Monitor must gracefully degrade to MV-from-optical-flow if sideband is absent.

4) EDID / Capability Advertisement (vendor block sketch)

Tag: IMUF
Fields:

upscale_modes: {1440p→4K, 1080p→4K}

framegen_multipliers: {×2, ×3, ×4}

max_reprojection_latency_ms: e.g., 1.0

bfi_support: {off, low, med}

mv_sideband: {required, optional, none}

vrr_passthrough: {yes/no}

color_depth_supported: {8/10/12bpc}

optical_flow_fallback: {yes/no}

Driver toggles PTM if IMUF is present and chosen in OSD or control panel.

5) Quality/Perf Targets

Motion clarity: ≥90% blur reduction vs 240 Hz baseline (with BFI on).

Inst. latency add: ≤1.2 ms at 4K×1000 Hz target.

Jitter/tearing: none; VRR-safe.

Failure modes: clear OSD status if sideband missing, flow confidence low, or reprojection disabled.

6) Prototype Path

PC-side PoC: two-thread render (native) + reprojection (visible) to prove lagless feel.

Firmware port: move upscale/reprojection to monitor SoC/NPU; wire up sideband; expose EDID IMUF.

A/B tests: artifact sweeps (pan/scroll/rotation), OD tuning, BFI comfort.

Hello.

If we say "kilo(s)", we write ALWAYS "k" - this letter must be lowercase due to the law for avoid any confusion:
- learn: https://www.nist.gov/pml/owm/metric-si-prefixes and more because "K" is Kelvin (temperature).

If we say "hertz", "HZ" and "hz" are incorrect for highercase by "Z" and for lowercase by "h" but yes we write always Hz.

Thank you.

The elephant in the room: GtG

The first 1000 Hz screen is likely to be an LCD, apparently.

However, GtG is already a giant percent of refreshtime. 1ms GtG is already 50% of the refreshtime of 500Hz (1/500sec = 2ms) and thus can more-or-less double the motion blur.

Thusly, a 500fps 500Hz OLED roughly equalling 1000fps 1000Hz LCD in display motion blur at framerate=Hz.

Chief Blur Buster wrote: ↑

23 Sep 2025, 22:58

The elephant in the room: GtG

The first 1000 Hz screen is likely to be an LCD, apparently.

However, GtG is already a giant percent of refreshtime. 1ms GtG is already 50% of the refreshtime of 500Hz (1/500sec = 2ms) and thus can more-or-less double the motion blur.

Thusly, a 500fps 500Hz OLED roughly equalling 1000fps 1000Hz LCD in display motion blur at framerate=Hz.

There have been some breakthroughs with LCDs recently.
Hong Kong University had some breakthrough with Ferroelectric LCDs
22 microseconds gtg at 5V, 9000:1 contrast ratio, field sequential color
https://sid.onlinelibrary.wiley.com/doi ... /jsid.2072
You may have ran into these guys at Display Week 2025


Blue Phase Mode LCD by University of South Carolina
https://www.sc.edu/uofsc/posts/2025/09/ ... ystals.php

So it seems that LCDs are catching up
Although of course no zero blacks without dimming zones or dual layers

Yes, but I'm worried about the optics (reputation)

- First 1000 Hz LCD comes out
- It's not one of these faster LCDs
- The GtG is still big part of refreshtime
- The predictable news articles about 1000Hz not being human-visible to eyes (without knowing nuanced context such as geometrics AND GtG=0)

Chief Blur Buster wrote: ↑

07 Oct 2025, 14:53

Yes, but I'm worried about the optics (reputation)

- First 1000 Hz LCD comes out
- It's not one of these faster LCDs
- The GtG is still big part of refreshtime
- The predictable news articles about 1000Hz not being human-visible to eyes (without knowing nuanced context such as geometrics AND GtG=0)

Personally I wouldn't worry about it because:

-Even with OLED you aleady have some of these kind of luddite articles and people who just like to complain technological improvements for no reason whatsoever

-LCD fab is so ahead of other technologies... whenever this kind of tech gets rolled out it will most likely be used even in entry-level $300 panels, and entry-level is where true standardization occurs... because when this happens suddenly every LCD panel on the market will now have response times as good as OLED or even better which is a huge jump in display standards, and the fact that this tech will most likely come at an affordable price point means that there won't be that many of these kinda cynical news articles because they won't have much to complain about

thatoneguy wrote: ↑

20 Oct 2025, 08:51

Personally I wouldn't worry about it because:

-Even with OLED you aleady have some of these kind of luddite articles and people who just like to complain technological improvements for no reason whatsoever

That's the UFO's job to worry about. As a honorary Hz mythbuster it's something I cannot avoid being mindful of;

It's still one of Blur Busters' raison d'etre to exist to educate mainstream media too, and there's been successes over the years.

It is especially powerful if Blur Busters / TestUFO educates just one media outlet at a time -- because there's a multiplier effect (their readers, who are typically not as Hz-educated as other fanbases). All those battles, sometimes I can't win all battles -- but I have to be wary and on guard for the potential amplifying of myths. One site says it's worthless, and then multiple sites starts to acknowledge, and then so on. It's happened multiple times in the past as Hz progresses.

TestUFO was invented partially as an easy way "educate in a shared link" -- by users and mainstream media about refresh rate physics. They still tend to use TestUFO as the gold standard.

Most recently yesterday when Mythbuster Adam in Adam Savage Tested Youtube Channel used TestUFO to confirm Apple Vision Pro M5's in-headset Safari browser is already 120Hz by default. (x.com | bsky).

Tech is constantly improving and any fast-response panels will produce quite distinct Hz differences compared to past LCDs, but the point still stands -- 1000Hz is going to arrive with a non-zero GtG. Maybe I can get some article views from a preemptive "1000Hz Can Either Be Worthless Or Amazing" bait article title I post before CES 2026. Media doing fact checking will see those first. Not all media does fact checking, but it's one of the things Blur Busters plays a role as a source/citation in fact checks.

It's hard to cover all types of media (youtube, vloggers, blogs, reviewers, users, social media, industry/science journals like Society for Information Display, etc) but the world is slightly more Hz-intelligent thanks to the UFO's role, sometimes preemptive education. Whether it's like a tweep who asks the memetastic "Grok, explain this" or some professional citation checking, "We verified this with these sources."